We've essentially gone through all the layers of the AI stack and have found different vulnerabilities. Our point is that anything you do with any one of these different vulnerabilities.... Some of them are worse than others, but you don't have to solve them all at the same time. You can work at mitigating exposure. On cloud, for example, building sovereign compute for the most highly sensitive national security is one way of making sure we don't get exposed to that type of vulnerability.
